JZTXT

窗口函数 mysql 8.0版本前实现方式

发布时间 2023-03-31 23:04:41作者: 何苦->

窗口函数 mysql 8.0版本前实现方式

MySQL5.7.25和 8.0.16 环境中实现类似Oracle的分析函数(8.0版本中已支持，直接使用即可)。

CREATE TABLE test1 (
	cus_no VARCHAR ( 10 ) DEFAULT NULL COMMENT '客户编号',     -- 客户编号 
	age INT ( 4 ) DEFAULT NULL COMMENT '年龄',                 -- 年龄 
	tran_month VARCHAR ( 6 ) DEFAULT NULL COMMENT '交易月份',  -- 交易月份 
	tran_date datetime DEFAULT NULL COMMENT '交易日期',        -- 交易日期 
	tran_amt DECIMAL ( 20, 2 ) DEFAULT NULL COMMENT '交易金额' -- 交易金额 
);
-- 插入测试数据
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101028',28,'201910','20191012',880.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101028',28,'201910','20191013',69.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101028',28,'201910','20191014',128.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101028',28,'201910','20191015',12.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101028',28,'201910','20191016',99.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101028',28,'201910','20191018',199.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101028',28,'201910','20191020',28.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101028',28,'201911','20191101',39.00);

insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101029',null,'201910','20191012',33.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101029',null,'201910','20191013',28.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101029',null,'201910','20191014',120.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101029',null,'201910','20191015',230.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101029',null,'201910','20191016',129.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101029',null,'201911','20191102',321.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101029',null,'201911','20191103',25.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101029',null,'201911','20191104',89.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101029',null,'201911','20191105',60.00);

insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101030',35,'201910','20191015',260.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101030',35,'201910','20191016',320.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101030',35,'201910','20191016',180.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101030',35,'201910','20191017',100.00);
insert into test1(cus_no,age,tran_month,tran_date,tran_amt) values('cus_101030',35,'201910','20191018',40.00);

模拟分区子句的使用

需求：统计每个客户在保留原来数据行的基础上增加一列统计列（消费总额）

例：

客户编号 年龄 交易月份 交易日期 交易金额 消费总额(追加列) 每个月的消费总额(追加列)

SELECT
	t1.cus_no '客户编号',
	t1.age '年龄',
	t1.tran_month '交易月份',
	t1.tran_date '交易日期',
	t1.tran_amt '交易金额',
	( SELECT sum( tran_amt ) FROM test1 WHERE cus_no = t1.cus_no GROUP BY cus_no ) AS '消费总额',-- 每个客户的消费总额
	( SELECT sum( tran_amt ) FROM test1 WHERE cus_no = t1.cus_no AND tran_month = t1.tran_month GROUP BY cus_no, tran_month ) AS '每个月的消费总额' -- 每个客户每个月的消费总额
FROM
	test1 AS t1;

模拟排序子句的使用

需求：统计每个客户逐日累计的消费金额

例：

2019-10-15当日消费金额 = 0+当日

2019-10-16当日消费金额 = 2019-10-15当日消费金额 + 当日

2019-10-17当日消费金额 = 2019-10-16当日消费金额 + 当日

以此类推。

SELECT
	t11.cus_no '客户编号',
	t11.fmt_date '交易日期',
	t11.total_amt '交易金额',
	sum( t12.total_amt ) '逐日累计消费金额'
FROM
	-- 按客户编号及交易日期进行合计分组生成基础数据
	( SELECT cus_no, sum( tran_amt ) total_amt, DATE_FORMAT(tran_date, '%Y%m%d') fmt_date FROM test1 GROUP BY cus_no, DATE_FORMAT(tran_date, '%Y%m%d') ) AS t11
	INNER JOIN 
	( SELECT cus_no, sum( tran_amt ) total_amt, DATE_FORMAT(tran_date, '%Y%m%d') fmt_date FROM test1 GROUP BY cus_no, DATE_FORMAT(tran_date, '%Y%m%d') ) AS t12 
	ON t11.cus_no = t12.cus_no AND t11.fmt_date >= t12.fmt_date
GROUP BY t11.cus_no, t11.fmt_date
ORDER BY t11.cus_no, t11.fmt_date

分析

SELECT
	t11.cus_no 't11客户编号',
	t11.fmt_date 't11交易日期',
	t11.total_amt 't11交易金额',
	t12.fmt_date 't12交易日期',
	t12.total_amt 't12交易金额'
-- 	sum( t12.total_amt ) '逐日累计消费金额',
FROM
	-- 按客户编号及交易日期进行合计分组生成基础数据
	( SELECT cus_no, sum( tran_amt ) total_amt, DATE_FORMAT(tran_date, '%Y%m%d') fmt_date FROM test1 GROUP BY cus_no, DATE_FORMAT(tran_date, '%Y%m%d') ) AS t11
	INNER JOIN 
	( SELECT cus_no, sum( tran_amt ) total_amt, DATE_FORMAT(tran_date, '%Y%m%d') fmt_date FROM test1 GROUP BY cus_no, DATE_FORMAT(tran_date, '%Y%m%d') ) AS t12 
	ON t11.cus_no = t12.cus_no AND t11.fmt_date >= t12.fmt_date
WHERE t11.cus_no = 'cus_101028'  -- 只看当前客户编号 cus_101028 的数据
-- GROUP BY t11.cus_no, t11.fmt_date -- 去掉取合分组 分析连表数据
ORDER BY t11.cus_no, t11.fmt_date

内联接 按t11.cus_no, t11.fmt_date排序，t11.cus_no, t11.fmt_date分组，

联接条件t11.fmt_date >= t12.fmt_date

模拟窗口子句的使用

需求：统计每个客户每个交易日期及其前后一天的消费总额

例：

日期 2019-10-15、2019-10-16、2019-10-17

2019-10-15 消费总额 = 0 + 2019-10-15 消费总额 + 2019-10-16 消费总额

2019-10-16 消费总额 = 2019-10-15 消费总额 + 2019-10-16 消费总额 + 2019-10-17 消费总额

2019-10-17 消费总额 = 2019-10-16 消费总额 + 2019-10-17 消费总额 + 2019-10-18 消费总额

以此类推。

SELECT
	t1.cus_no '客户编号',
	t1.age '年龄',
	t1.tran_month '交易月份',
	t1.tran_date '交易日期',
	t1.tran_amt '交易金额',
	( SELECT sum( tran_amt ) FROM test1 
		WHERE cus_no = t1.cus_no 
		AND date_format(tran_date, '%Y-%m-%d') >= date_sub(date_format(t1.tran_date, '%Y-%m-%d'), interval '1' day) 
		AND date_format(tran_date, '%Y-%m-%d') <= date_add(date_format(t1.tran_date, '%Y-%m-%d'), interval '1' day) 
		GROUP BY cus_no
	) AS 'range消费总额',-- 统计每个客户每个交易日期及其前后一天的消费总额
	( 
		SELECT sum(t_11.total_amt) FROM (
			-- <2> 把处理好的数据增加一个行号 rownum
			SELECT *, ( @row1 := @row1 + 1 ) AS rownum FROM (
				-- <1> 把所有数据按客户编号 交易日期进行排序
				SELECT *, date_format( tran_date, '%Y-%m-%d' ) AS fmt_date, sum(tran_amt) total_amt 
				FROM test1 
				GROUP BY cus_no, date_format( tran_date, '%Y-%m-%d' ) 
				ORDER BY cus_no, date_format( tran_date, '%Y-%m-%d' ) 
				-- <1> 
			) t_1, ( SELECT @row1 := 0 ) AS t 
			-- <2> 
		) t_11
		INNER JOIN (
			SELECT * FROM (
				-- <2> 把处理好的数据增加一个行号 rownum  注意：row2的变量名不能与上面的相同
				SELECT *, ( @row2 := @row2 + 1 ) AS rownum FROM (
					-- <1> 把所有数据按客户编号 交易日期进行排序
					SELECT *, date_format( tran_date, '%Y-%m-%d' ) AS fmt_date, sum(tran_amt) total_amt 
					FROM test1 
					GROUP BY cus_no, date_format( tran_date, '%Y-%m-%d' ) 
					ORDER BY cus_no, date_format( tran_date, '%Y-%m-%d' ) 
					-- <1> 
				) t_1, ( SELECT @row2 := 0 ) AS t 
				-- <2> 
			) t_12 
		) t_13 ON t_13.cus_no = t_11.cus_no 
		WHERE
			-- t1表参数只能放在最外层使用
			t_11.rownum >= t_13.rownum - 1 AND t_11.rownum <= t_13.rownum + 1 and t_13.cus_no = t1.cus_no and t_13.fmt_date = date_format(t1.tran_date, '%Y-%m-%d')
) AS 'rows消费总额' -- 统计每个客户每个交易日期及其前后一条的消费总额
FROM
	test1 AS t1;